The point of my original comment was precisely that SGML was optimized for text documents. I agree that adopting it for configs was a mistake, but the complaint that "IBM lawyer who invented the SGML syntax had never heard of S-expression" doesn't make any sense in that context.